Graphcast: Learning Skillful Medium-Range Global Weather Forecasting

By Remi Lam et al

Published on Aug. 4, 2023

Read the original document by opening this link in a new tab.

Table of Contents

Keywords: Weather forecasting, ECMWF, ERA5, HRES, learning simulation, graph neural networks
Introduction
GraphCast
Verification methods
Forecast verification results
Severe event forecasting results

Summary

Global medium-range weather forecasting is critical for decision-making in various domains. Traditional numerical weather prediction methods lack the ability to improve accuracy with historical data. In this paper, GraphCast, a machine learning-based method, is introduced for accurate and efficient weather forecasting. It outperforms traditional systems on 90% of verification targets, supporting better severe event prediction. GraphCast utilizes deep learning techniques and shows promising results in medium-range forecasting. The model is trained on historical data and achieves impressive results in comparison to existing operational systems like HRES. The study evaluates the performance of GraphCast in predicting severe events like tropical cyclones and atmospheric rivers, demonstrating its superiority over traditional methods.